The Method for Prediction the Distribution of Information in Social Networks Based on the Attributes

Social networks are important part of society. Analysis of information flows and activity of users may provide different opportunities for scientist. However, existing methods and models for prediction of information flow couldn’t provide sufficient level of accuracy. Proposed method of information flow analysis is based on machine learning algorithms. First stage of research aims to analyze the features, that could be useful for describing activity of users and information in social networks. Data was assembled from social network “VKontakte”. Data includes main information about selected features for each person. On the second stage were conducted experiments, based on machine learning methods. Three main approaches were implemented in research - naive Bayes classifier, logistic regression analysis and k-means clustering. A comparison of different machine learning methods was conducted on last stage. Best results of predictions were reached by usage of naive Bayes classifier. Proposed method for prediction of information flow is usable for organizing users’ protection from different type of information attacks.